Financial markets are inherently non-stationary, with interest rate environments, regulatory policies, market participant structures, and macroeconomic cycles continuously evolving, which violates the assumption in traditional supervised learning that training and test data are identically distributed.